Dissertation Discussion and Conclusion Chapters: Worked Examples (UK, 2026)

By the time you reach the discussion you have been writing for months, the deadline is close, and the temptation is to restate the results in different words and stop. That is the single most expensive shortcut in an undergraduate dissertation, because the discussion is where critical judgement becomes visible — and criticality is what separates the upper degree classifications.

This guide gives you the five moves of a discussion chapter and the four jobs of a conclusion, each with a fully written worked example you can hold your own draft against.

The line between results and discussion

The results chapter says what the data show. The discussion says what that means. In practice the boundary is a grammatical one: results sentences describe, discussion sentences explain, compare and qualify.

If a sentence in your results chapter contains “suggests”, “because”, “consistent with” or “may reflect”, it is a discussion sentence sitting in the wrong chapter. Our guide to the results chapter covers the reporting conventions on the other side of that line.

The five moves of a discussion chapter

1. Answer the research question directly. One short paragraph, in plain words, before anything else.
2. Interpret each finding against the literature. Where it agrees, where it disagrees, and why.
3. Explain the unexpected. Null results and surprises get more space than confirmations, not less.
4. State the limitations and what each does to your claims.
5. Draw out the implications — for theory, for practice, and for the next study.

A worked discussion section

This is move 2 written out in full, for the hybrid-working study used throughout our guides. Read it for the pattern rather than the content.

The positive association between hybrid autonomy and retention intention is consistent with the large-organisation literature reviewed in Chapter 2, and its presence in firms without formal HR functions is the study’s main contribution. Job embeddedness theory holds that employees stay because of accumulated links to the organisation; where formal policy does not exist to create those links, the present finding suggests discretion itself may serve the same function.

The effect was, however, smaller than that reported in comparable large-firm studies. Two explanations are available from the data. The first is measurement: autonomy was captured here as days of location choice, a coarser indicator than the multi-item scales used elsewhere, and coarser measurement attenuates observed associations. The second is substantive: in small firms, informal arrangements may be less stable, so a given level of autonomy carries less assurance that it will persist. The present design cannot distinguish these, and doing so would require a measure of perceived permanence, which is a concrete direction for further work.

The moderation hypothesis was not supported. Tenure did not significantly qualify the relationship, contrary to expectation, though the achieved sample gave limited power to detect an interaction of the size the literature implies. The result is therefore better read as inconclusive than as evidence of absence.

Three things that paragraph does deliberately. It states which finding is the contribution rather than leaving the reader to work it out. It offers two competing explanations and admits the design cannot choose between them. And it refuses to treat a non-significant result as proof of no effect, which is the most common inferential error in undergraduate discussions.

What to do when the results are not what you expected

Most students panic. Markers do not. An unexpected or null result is entirely normal empirical work and is marked on how you handle it, not on whether it occurred.

Work through four explanations in order, and say which you find most plausible:

1. The effect is genuinely absent in this population or setting — which is a finding.
2. The measurement was too coarse to detect it.
3. The sample was too small to detect an effect of the expected size.
4. A moderator was operating that the design did not capture.

What you must not do is describe a non-significant result as “approaching significance” or quietly reframe your hypothesis to match the data. Changing a hypothesis after seeing results, without saying so, is a serious problem rather than a stylistic one.

The conclusion’s four jobs

A conclusion is short — often 500 to 800 words at undergraduate level — and it does exactly four things:

1. Answers the research question, in a sentence each.
2. States the contribution: what is now known that was not known before.
3. Gives recommendations, each one traceable to a finding.
4. Closes with a forward statement about the next study.

A worked conclusion

This dissertation asked how hybrid working arrangements influence retention intention in UK SMEs, and whether the degree of autonomy an arrangement confers accounts for any relationship. It found a positive association between location autonomy and retention intention among employees of firms with fewer than 250 staff, weaker than that reported in large-organisation studies, and not moderated by tenure in this sample.

The contribution is contextual rather than theoretical. Job embeddedness has been examined almost exclusively where hybrid working is formal written policy; the present study provides initial evidence that the relationship persists where arrangements are negotiated individually, which suggests the mechanism does not depend on formal policy to operate.

Two recommendations follow for owner-managers. Arrangements that give employees genuine choice over location appear more closely related to retention intention than arrangements that merely permit remote working on fixed days, which implies that the discretion is the active ingredient rather than the days. And because informal arrangements were more varied within firms than between them, stating the arrangement explicitly — even without a formal policy document — is likely to matter.

The obvious next study is longitudinal, measuring autonomy before a subsequent retention outcome, with a measure of perceived permanence included. Until that exists, the relationship reported here should be read as an association rather than an effect.

Notice that nothing new appears. Every claim traces back to a finding already reported, and every recommendation names who would act on it.

Should the discussion and conclusion be one chapter or two?

Both are common in UK departments and your handbook decides. If they are merged, the risk is that the answer to the research question gets buried in the middle of a long chapter; guard against it by opening the merged chapter with the direct answer. If they are separate, the risk is repetition; guard against it by keeping interpretation in the discussion and restricting the conclusion to answer, contribution, recommendations and next steps.

Tense shifts across these chapters in ways that trip people up — our guide to choosing the right tense for each chapter covers where UK guidance genuinely disagrees. How the limitations paragraph is constructed is covered in our guide to scope and limitations, and where these chapters sit in your project’s overall shape is in our guide to dissertation structure by subject.

Five mistakes markers flag

1. Restating results without interpreting them. The commonest cause of a discussion that reads as descriptive.
2. New data or new citations in the conclusion. Everything in the conclusion should already have appeared.
3. Treating non-significance as proof of no effect. Say “did not detect”, not “showed there is no”.
4. Recommendations that float free of findings. Each one should be traceable to a specific result.
5. No direct answer to the research question. A marker should find it in the first paragraph of the discussion and again in the conclusion.

If the analysis is finished and these two chapters are what stand between you and submission, that is the most common place to stall. Tesify drafts your discussion and conclusion from the findings and literature you have already entered, so you are editing an argument rather than starting one. Everything stays 100% written by you, and it is free to start.

Frequently asked questions

How long should the discussion chapter be?

Commonly 20–25% of the total word count, so roughly 2,000–2,500 words in a 10,000-word dissertation. The conclusion is much shorter, typically 500–800 words.

What is the difference between the results and the discussion?

Results report what the data show, without interpretation. The discussion explains what it means, compares it with the literature and qualifies it. Interpretive verbs in the results chapter are the tell that the boundary has been crossed.

Can I include new literature in the discussion?

A little, where an unexpected finding needs explaining from work you had no reason to review earlier. Substantial new literature suggests the review missed something and should have been in Chapter 2.

What do I do if my hypothesis was not supported?

Report it plainly, work through the four explanations — genuine absence, measurement, power, unmeasured moderator — and say which you find most plausible. Do not change the hypothesis retrospectively.

Should the conclusion repeat the findings?

It should state the answers, briefly, not repeat the reporting. One sentence per research question is right.

Where do the recommendations go?

In the conclusion in most UK dissertations, though some departments place them at the end of the discussion. Either way, each recommendation must trace to a finding.

Do I need a separate implications section?

Only if your handbook asks. Otherwise implications for theory and practice sit as the final move of the discussion, immediately before the limitations or immediately after them.

How do I avoid repeating myself across the two chapters?

Give each a distinct job. Discussion interprets and qualifies; conclusion answers, states the contribution and recommends. If a paragraph could sit in either, it belongs in the discussion.